Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add rules/prisma/orm/non-vacuous-verificationgit clone --depth 1 https://github.com/prisma/ormWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00342 |
| Opus 5 | $0.00000 | $0.00171 |
| Sonnet 5 | $0.00000 | $0.00068 |
| Haiku 4.5 | $0.00000 | $0.00034 |
Grade A, and why
non-vacuous-verification scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Non-vacuous verification
A verification that cannot fail converts absence of evidence into false confidence. Before trusting a green result, name what would have made it red.
- A regression test must be red before the fix — run it against the unfixed code and record the failure. If the triggering scenario cannot be reproduced (e.g. a transport race), don't ship a test that is green in both worlds: make the seam testable and test the seam directly.
- A verifier must fail (or loudly warn) on empty input. "All 0 known cases verified" is a discovery failure, not a pass.
- Guard type tests against
anypoisoning.skipLibCheck+ an unresolvable import in a fixture.d.tsmakes every dependent assertion vacuous (T & anyisany). Add anot.toBeAny()sentinel on the fixture's central types. - Comments are not evidence. A comment describing a guard does not mean it exists — read the code it claims to describe; correct the comment when it lies.
- Generated files: test claims by regenerating. Reading the generator is not evidence; running it is.
- Instrument before designing a fix for a hard-to-reproduce failure — a counter proving "this catch never runs" kills a wrong hypothesis in one run.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 21 lines · 0 tokens per session scan A 17c711dcf096
non-vacuous-verification is a cursor rule published in the GitHub repository prisma/orm (47,587 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 342 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other cursor rules, from other repositories
playwright-e2e
Playwright E2E testing configuration for AI agents.
react-components
Rules for writing React components in this project.
state-management
Rules for Redux state management.
server-patterns
Rules for server-side Express code.
project
AI Phone Agent — TypeScript voice backend (Twilio + Amazon Connect).
typescript-style
TypeScript style — prefer arrow functions and consistent callbacks.